Localization of the NMR signal of a tissue or an organ of interest is a ma jor problem of in vivo NMR spectroscopy. Many localization techniques exist (I-5) but each has its drawbacks and artifacts (eddy currents, delays, T2, and chemical-shift effects). Those conceptually most appeal ing are relatively demanding of hardware, often suffer from technological or practical lim itations, are not always easy to implement, and are, therefore, prone to a number of errors. This is, for instance, true of mu ltiple-shot subtraction techniques such as ISIS, which is largely used for 3’P spectroscopy, where movement or pulse imperfection may lead to large subtraction errors. We here suggest a simple “biochemical” method of correcting for the contamination of the desired signal by that of another tissue. We first used this method in 1984 when not many localization techniques were available; the motivation to write it down came from a recent meeting at which people were worrying about the residual phosphocreatine (PCr) peak present in their liver or kidney spectra, despite localization in one or more dimensions. The method is applicable when at least one peak may be attributed to the contaminating tissue only and when the relative peak areas of the contaminating tissue are known. This is, for instance, the case with liver observation by a surface coil, where the “P spectra can be contaminated by signals from nearby muscle, as reflected by the presence of PCr in the spectra. Since liver lacks PCr, all the PCr in the spectra can be attributed to muscle (Fig. 1). One should, of course, initially try to m inimize signal contamination as judged by the relative height of the PCr peak, for example, by using a 180” pulse at the coil center rather than 90” or less (Fig. 2) or by more sophisticated “depth” pulses (6), which already give relatively good suppression of PCr, i.e., of contamination. To a first approximation, one could then simply ignore the contamination reflected by the small PCr signal, since other muscle signals are at least four times smaller (Fig. 3b) ( 7). To a second approximation, one can subtract a muscle spectrum (or corresponding peak areas) from the observed spectrum (Fig. 1). How much muscle spectrum should be subtracted is given by the requirement of canceling the PCr peak as there is none in the liver. 3’P spectra were acq uired from a normal subject at 2 T with a 3 or a 10 cm diameter surface coil. A simple pulse-acquire sequence was used with a nearly nonsaturating repetition time (3 s). The pulse angle at the coil center is specified where relevant. For comparison, liver spectra were obtained using a 200 ps surface-spoiling gradient pulse, applied during the dead time before acquisition as described elsewhere (8).